[PATCH] D105996: [AArch64] Enable Upper bound unrolling universally

Thu Jul 15 03:51:55 PDT 2021

jaykang10 added a comment.

In D105996#2877692 <https://reviews.llvm.org/D105996#2877692>, @SjoerdMeijer wrote:

> The perf data looks a bit thin, and I guess this is for 1 cpu, the N1? Not sure if we need some more data (and different core) to confirm this is a good thing? Or do we think this is enough?

The performance data from spec2006 for AArch64

  Benchmark		diff (%)
  400.perlbench		0.346187575
  401.bzip2		0.552000394
  403.gcc		-0.4505745
  429.mcf		-2.075237465
  445.gobmk		0.091198338
  456.hmmer		0.010805569
  458.sjeng		-0.063621626
  462.libquantum		-3.733043957
  464.h264ref		0.212556072
  471.omnetpp		-2.088510101
  473.astar		-0.312109961
  483.xalancbmk		-0.699945194

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D105996/new/

https://reviews.llvm.org/D105996